IMPLEMENTASI MODEL TRANSFORMERS T5-small UNTUK TEXT SUMMARIZATION DALAM PEMBUATAN META DESKRIPSI SEO

The manual creation of meta descriptions for Search Engine Optimization (SEO) is an inefficient, time-consuming process that struggles with consistency, especially for websites with a large volume of content. This research aims to implement and evaluate a Transformer-based text summarization model, specifically the T5-small variant, to automate the generation of SEO meta descriptions for Indonesian language content. The research method involves further fine-tuning the panggi/t5-small-indonesian-summarization-cased checkpoint model by retraining it on the large-scale news dataset fajrikoto/id_liputan6. This model adaptation process was conducted in the Kaggle Notebooks cloud computing environment. The model's performance was measured using standard NLP evaluation metrics, namely ROUGE and BLEU. Quantitative evaluation results on 10,972 test data samples show promising performance, achieving a ROUGE-1 score of 34.09, ROUGE-2 of 18.67, ROUGE-L of 29.37, and a BLEU score of 10.41. The fine-tuned model was then successfully implemented into a functional web application using the Streamlit framework as a practical proof of concept. This study demonstrates that the T5-small model has the potential to be a solution for improving efficiency and scalability in SEO content production. It also contributes to the application of NLP in the digital marketing domain in Indonesia.
Keywords: text summarization, Transformer T5-small, Meta Description SEO, Search Engine Optimization (SEO), T5-small, Natural Language Processing (NLP).